Please report new issues athttps://github.com/DOCGroup
Storing this publicly, have to discuss this with some people first In the generated stubs we have the following each time: if (!this->is_evaluated ()) { ::CORBA::Object::tao_object_initialize (this); } if (this->the_TAO_Base_Proxy_Broker_ == 0) { Base_setup_collocation (); } The setup collication, isn't this something what could be done within the !is_evaluated part? When no collocation is possible, why do we do it each time? Related also to this, at the moment direct and through collocation has been suppressed in the IDL compiler (-Sp and -Sd), do we then need to generate the setup_collocation methods? We don't want to use collocation, so why do we generate the setup call?
Reported by Johnny but forgot to logout first
fixed summary
The dummy TAO::Collocation_Proxy_Broker pointer value "0xdead" should probably be changed so that something like a "null" proxy broker object (not a pointer!) is used instead. This is necessary to prevent a crash under configurations where direct collocation is disabled in the TAO_IDL generated skeletons but selected as the collocation strategy by the user at run-time. Here's the ChangeLog that explains why the 0xdead dummy value was introduced: Tue Mar 8 22:08:49 2005 Ossama Othman <ossama@...> ... * TAO_IDL/be/be_visitor_interface/interface_sh.cpp: * TAO_IDL/be/be_visitor_interface/interface_ss.cpp: Only generate a strategized proxy broker for the direct collocation case. A strategized proxy broker isn't needed for the thru-POA collocated case. Reduces skeleton footprint by approximately 2.5Kb per interface. Generate a dummy collocation proxy broker factory function in the thru-POA collocation case that returns a dummy TAO::Collocation_Proxy_Broker pointer (i.e. "0xdead"). The collocation code in TAO's ORB_Core still detects collocation availability by checking for a non-zero pointer. That value will not be used for the thru-POA case. A valid collocation proxy broker factory function is still generated for the direct collocation case.
It turns out the 0xdead issue is already captured in bug 2241.
Seems more work on this, Ossama, any ideas on moving the setup_collication into the first if?
can't move the call into the if, closing this one